Pancreas: Tumors: Imaging Pancreatic Cancer: The Role of Multidetector CT (MDCT) with 3D CT Angiography

Elliot K. Fishman, MD, Karen M. Horton, M.D.

Introduction

Computed Tomography (CT) remains the radiological study of choice in most institutions for the evaluation of suspected pancreatic disease (1-4). Although other modalities such as MRI, PET and Endoscopic Ultrasound have their proponents, CT continues to make significant technological advancements which support it as the leading imaging modality for pancreatic pathology. These new CT techniques promise to revolutionize pancreatic imaging and improve patient care and management. To fully understand the current role of CT, it is helpful to review the evolution of CT scanning

The Evolution of CT Scanning

With the introduction of Computed Body Tomography (CT) into clinical practice in the late 1970’s, it was finally possible to directly visualize the pancreas non-invasively. Prior to that point, imaging of the pancreas was poor and limited to early generation ultrasound, nuclear medicine studies and barium studies (upper GI series with barium). These studies only inferred the presence of disease by secondary signs. When these studies were positive the patient typically had a large tumor which was not amenable to curative surgical intervention. More invasive techniques such as angiography were sometimes utilized but were limited to vascular maps and could not image the pancreas directly. CT provided a non-invasive technique which could directly visualize the gland appearance including the texture of the gland, the presence and extent of inflammation, the presence and extent of a pancreatic mass. In addition, CT could assess key peripancreatic vascular structures including the celiac axis, the superior mesenteric artery and vein and the portal vein. Additionally, important organs like the liver and peripancreatic lymph nodes could be viewed during a single non-invasive examination.

Today, over twenty years later, CT remains the primary study for evaluating the pancreas. However, significant changes in CT technology and the introduction of three-dimensional imaging make CT imaging of the pancreas significantly different than in those early days. Although a detailed analyses of CT scanner technology is beyond the scope of this article, it is important to understand the advancements that have occurred in order to truly appreciate the potential role of CT today in imaging pancreatic pathology. For instance, CT imaging of pancreatic cancer can improve patient outcome by correctly detecting not only tumor but also by evaluating the extent of disease which determines which patients could benefit from aggressive surgical management. Statistics as to the accuracy of CT as defined in the published literature must be viewed in light of the scanner technology used at the time the article was published. Unfortunately well-controlled clinical trials take several years to complete, and in an era of rapid technological change, are outdated by the time they are published.

In the early 1980’s, the state of the art scanners took approximately 10 seconds per slice to acquire the data, and approximately 60 seconds to reconstruct each image. Up to 4 slices could be obtained per minute and a typical scan collimation was either 5 or 10 mm. In order to scan the pancreas and liver an average scan time would be in the range of 5-6 minutes. This had obvious limitations as the contrast bolus had a peak enhancement of less than a minute which resulted in many articles and discussions as to selecting the ideal combination of scan protocol and contrast injection. In the early years of CT the contrast used was ionic and contrast reactions were common, thereby creating a different set of problems. Although most reactions were minor (i.e. nausea and vomiting) they often interrupted the scanning which further diminished the image quality. In the mid 1980’s the next generation of scanners was introduced that provided faster scan times (2-3 seconds per slice) and shorter reconstruction intervals (less than a second per slice on some scanners) with higher scan resolution. Scan slice thickness also improved and collimation in the 2-5 mm range became more common. Another major improvement was the introduction of non-ionic contrast, which decreased the incidence and severity of minor contrast reactions. Initial acceptance of nonionic contrast agents was limited by cost. Eventually through a combination of decreasing costs and the recognized advantages of these agents, their use is now nearly universal.

In the last years of the 1980’s, the most revolutionary concept since the introduction of CT, spiral CT was introduced. Spiral CT combined the need for fast scanning and optimal use of IV contrast material. Unlike classic CT where the patient was moved a preselected distance (i.e. 3-10 mm) and a scan done, followed by another table incrementation and another scan obtained (i.e. scanning and patient movement are two separate steps), the spiral scanner moved the patient through the x-ray beam continuously and acquired the data in this continuous fashion. The technology required a slip ring detector array as well as advances in data acquisition and processing. Although initial models of the scanner could only acquire 12 seconds of information, this soon gave way to 24 seconds, 36 seconds and eventually 60-second acquisition times.

The advantages of spiral CT are many, including the ability to acquire volumes of data rather than just slices of data. These volumes are acquired during a single breathold in contrast to non-spiral which requires 30-50 individual breatholds. Problems associated with patient motion, inconsistent breathholds or misregistration were eliminated or sharply reduced with spiral CT. The ability to acquire information as a volume also produced several unique capabilities. For example, if the abdomen were scanned, the user could retrospectively review the data at any chosen increment. In other words, if we scanned a volume of 20 cm we could review the data at 5mm intervals for a total of 40 scans or every 1-mm for a total of 200 scans. Previous work has shown that increased data sampling increases the lesion detection rate in the lung and liver. The increased data sampling was also ideal, as it improved data sets for three-dimensional reconstruction of CT data. One limitation of these spiral scanners was that the slice collimation had to be chosen prior to the scan and could not be changed. Selection of slice width is important for both lesion detection and data quality for 3D imaging. The compromise was that with narrow collimation the distance covered (volume scanned) was decreased. For example, a 30-second spiral scan using 5-mm collimation could cover 5 times the volume when compared to a 30 sec spiral scan using 1 mm collimation, if all other parameters were held constant.

The final technological advance came in the late 1990’s with the introduction of multislice or multidetector CT scanning (5-6). This innovation was built on the desire to acquire information faster, with better resolution and for larger scan volumes. Multidetector CT scanner varies among manufacturers but certain concepts are universal. Just like with spiral CT, the data acquisition is continuous but instead of 1 detector gathering information there are 4 detectors. In addition, the scanner's rotation speed is typically double (or 500 milliseconds) the rotation speed of a single detector system. In simple terms, it means that the multidetector scanner can acquire data up to 8 times faster than the single detector system. In addition, unlike single detector CT where slice width is fixed at time of scanning, the MDCT scanner allows the slice width to be changed retrospectively. For example, with 1 -mm collimation we can create slice scan widths ranging from 1 to 6 mm. The scans also have improved slice profiles compared to single detector CT and higher resolution (up to 24 line pairs vs. 16 line pairs for most single detector scanners). The ability to scan with high speed and high resolution allows optimal coupling of contrast injection and scan acquisition. This is of critical importance in imaging organs such as the pancreas and liver. In addition, the ability to acquire multiple acquisitions in a rapid sequence allows acquisition of images in multiple phases of contrast enhancement including arterial phase dominant scans, portal venous phase dominant scans, and equilibrium scans (7,8). A set of typical scan parameters comparing single and multidetector CT is listed in table 1.

The other major advance has been the development of CT based 3D image-processing workstations. Although they have been around since the early 1980’s the new systems are lower cost ($50-100,000 vs $120-250,000), are easier and faster to use, and provide better rendering algorithms including volume rendering. Volume rendering was a technique initially introduced by LucasFilms (San Rafael, CA) but when applied to medical imaging provided highly detailed accurate three-dimensional images. Unlike earlier techniques like shaded surface the rendering parameters can be varied interactively to display soft tissue, vessel and bone. The newest workstations take advantage of faster computer processing boards to provide a real time environment, which allowed the process of 3D imaging to be both interactive and fast to perform. We currently use a prototype Siemens 3Dvirtuoso Workstation (Siemens Medical Systems, Iselin, NJ) for our 3D imaging that is equipped with advanced processor boards.

The specific details of real time rendering are beyond the scope of this article but a few of the basic concepts are especially important in pancreatic vascular imaging. By looking at the entire CT data set as a volume it is possible to choose the optimal plane and orientation of each vessel and its branches whether it be the SMA or the SMV . Any narrowing or vessel compression can be displayed as well as collateral pathways. Differentiation of a tumor abutting a vessel vs invading it is easier to define on this volumetric display. In addition, for the referring surgeon the ability to view the CT data as a display similar to classic angiography is a valuable adjunct in surgical decision making. Displays of variants of normal vascular anatomy including accessory right hepatic arteries are best defined with this technique (9-12).

Scan Protocols

Although scan protocols will vary from institution to institution depending on the specific scanner, the protocols for MDCT at Hopkins include dual phase imaging with data obtained 25 and 60 seconds after the injection of 120 cc of Omnipaque -350 (Nycomed Amersham, Princeton, NJ) at a rate of 3 cc/second. Scan parameters are 1.25 scan width with the protocol details listed in table 1. All scans are performed on a Siemens Somatom Volume Zoom scanner (Siemens Medical Systems Inc, Iselin, NJ). After data acquisition the images are sent over the hospital-imaging network to the workstation and select 3D images are generated from both phases of acquisition. These images are then put on film and sent directly to the referring physician. More information regarding scan protocols as well as additional case material can be found on our site on the World Wide Web at www.ctisus.com.

Pancreatic Mass Detection

The classic description for a pancreatic tumor was a mass within the pancreas commonly ranging in size from 3-6 cm. Detection of the mass was typically based on size parameters with less attention paid to the differential enhancement of normal and abnormal pancreatic tissue. In part this was due to the slow injection rates as well as longer scan times on early generation CT scans. Over time, the improvement in CT scanners allowed smaller masses to be detected. Signs commonly used include the presence of pancreatic or common bile duct dilatation, as well as changes in the pancreatic contour. Although most pancreatic masses are best seen on the portal venous phase images, as they are hypovascular . Other tumors including islet cell tumors and metastases are hypervascular and are best seen on arterial phase images. Graf et al. found that tumor conspicuity of pancreatic adenocarcinoma was better in the portal phase where the tumor to pancreas contrast difference was 54+/-31 H than in the arterial phase when it was 31 +/- 29H (13). Others including Choi et al. (14) have found the arterial phase to be more valuable. However, that series used late phase imaging at 180 seconds which is not an optimal imaging time selection. Regardless of the tumor type, the use of thin section CT with close interscan spacing allows smaller tumor detection when changes in the gland enhancement patterns are detected. Although axial imaging and review of these images alone has been the standard mode of CT review, it would not remain the gold standard, as newer technologies became available. Bonaldi et al. (15) found that simply reviewing the images on a workstation with a cine display provided better definition of tumors as well as vascular anatomy and ductal anatomy. The use of 3D imaging can be very helpful in select cases by defining the mass or suspected mass in multiple viewing planes. In many cases this may distinguish a true pancreatic mass from adjacent duodenal or small bowel tumor or peripancreatic adenopathy. The use of dual phase imaging is often helpful with MDCT in analyzing the etiology of the tumor. However, with pancreatic adenocarcinoma the portal phase (50-70 seconds) remains ideal for lesion detection.

Hepatic Lesion Detection

Detection of liver metastases is crucial when staging cancer as it will typically make a patient ineligible for attempt at surgical cure. The accuracy of CT has varied in the past but with spiral CT approaches 90% for a 1cm lesion (16). This was with single detector CT using 5-mm thick sections reconstructed at 5-mm intervals. With MDCT and the use of 1.25-3 mm thick sections at 1-3 mm intervals this undoubtedly will improve detection. However, no article has yet been published with results for pancreatic cancer. The use of thin sections coupled with dual phase acquisition allows the detection of both hypovascular and hypervascular lesions. Metastases from vascular pancreatic tumors like islet cell tumors are often hypervascular and may only be seen on the arterial phase of data acquisition. Metastases to the liver from adenocarcinoma are typically hypovascular and are best seen on the portal venous phase of data acquisition. In fact although they may be seen on the arterial phase they may be difficult to detect and distinguish from other lesions including hemangiomas.

Vascular Mapping

The area of probably the greatest controversy over time in pancreatic imaging has been the ability of CT to accurately determine the presence of vascular invasion. The accuracy of CT to define arterial or venous invasion has been the subject of numerous articles often focusing on its accuracy when compared to either catheter based angiography or to surgical findings. Several articles even in the pre-spiral CT era has clearly shown the equivalence of carefully performed CT angiography and catheter angiography for vascular encasement (17-18).

Yet, with spiral and then multidetector CT our capabilities have gone far beyond looking at vessels in the axial plane. As shown in figure x there are significant limitations when looking at vessels in the axial plane. Partial averaging of data, questions as to a tumor being adjacent to rather than encasing a vessel as well as the lack of an ideal display are problems with axial CT alone. The use of image reconstruction whether it is multiplanar reconstruction of a 3D map has obvious advantages especially to the referring physician who is more comfortable with a volumetric display. Let us the look at this in more detail by evaluating both the critical arterial and venous anatomy (19-21).

Arterial Imaging

The accurate display of arterial anatomy (SMA, celiac axis) is the major focus of arterial phase 3D mapping in the evaluation of the pancreas. The goal of preoperative vascular mapping in patients who are potential candidates for a Whipple procedure is to clearly define the angiographic map with accuracy equal to a classic catheter angiogram. With 3D CT angiography this goal can be obtained. Anatomic variation such as a common celiac and SMA trunk or the presence and location of a right hepatic artery arising off the SMA are all easily detailed on the 3D display. The viewing of this data with a stereoscopic display may add further information in cases of complicated vascular anatomy.

In cases where there is potential vascular invasion the 3D display will define the course of the vessel and its relationship to the pancreatic mass . The ability to determine whether a mass actually encases or just abuts a vessel will be clearly shown in most cases. The ability to view images in any plane or perspective is an important tool supplementing the information from the axial CT and in replacing angiography. Although no large series has yet to be published, our personal experience has been suggestive of a near one to one correlation with surgical findings. A detailed analysis of our results is currently under way.

With axial images alone, numerous articles tried to develop strategies for defining resectability. Lu et al. (22) graded vessel invasion on a 0-4 scale based on circumferential contiguity of tumor to vessel. That is, a grade 0 was no contiguity of tumor to vessel, grade 1 was tumor contiguous with less than one quarter circumference, grade 2 between on quarter and one half circumference, grade 3 between one half and three quarters circumference and grade 4 greater than three quarters circumference or any vessel constriction. Involvement of more than one half the circumference or grade 3 was highly specific for unresectable tumor.

Yet, even half of the grade 2 cases were proven to be unresectable and 12.5% of grade 3 were resectable. Although these results were strong we believe they do highlight some of the disadvantages of viewing vessels with an axial plane only. This is especially true when it comes to vessel narrowing or constriction. The use of MDCT increases the number of slices through the pancreas over even the best-published single detector spiral protocol by a factor of 2 or 3. This increased data sampling at 1-1.25 collimation and 1 mm interscan intervals coupled with the higher resolution of MDCT should prove better at defining the key arterial vessels even if only the axial images are reviewed. However, the ability to view the vessels in multiple orientations provides a more complete and accurate display where even subtle vessel tapering can be detected. The extent of vessel encasement is optimally defined with this visualization. Recent articles by Horton et al and Fishman et al. have clearly demonstrated the superiority of this new technology (23-24).

Venous Imaging

Numerous articles including one by Vedantham et al. (25) have shown that helical CT performed in the portal venous phase (AKA pancreatic phase) which is 40-70 seconds after injection of contrast is ideal for defining peripancreatic venous anatomy for determining tumor invasion. These authors agree with our conclusions that later phase imaging (120-180 seconds) is suboptimal for determining venous invasion. However as noted previously the use of axial CT alone is not ideal for pancreatic imaging and that a more volumetric display is required. With the development of single detector and now multidetector CT we believe that a volume display is the most accurate technique for evaluating venous invasion. Even with the use of single detector CT Graf et al. (26) were able to create accurate CT venograms of the mesenteric veins that were able to equal angiography for defining variations in vascular anatomy. However careful analysis of the images would suggest that their detail was not adequate to define early vessel invasion as in pancreatic cancer. Novick et al. did show that using advanced 3D techniques like volume rendering those vascular maps could not only define the venous anatomy but also accurately predict vascular invasion. Raptopoulos et al. (27) had similar results with CT Angiography from single detector spiral CT data sets. They found that by adding the CT angiogram to the axial images alone the negative predictive value of a resectable tumor was 96% compared to 70% for axial images alone. The studies were especially valuable in determining unresectability.

We have recently shown that the use of MDCT coupled with three-dimensional imaging provides an even better way to image vessel patency . By acquiring data sets of narrow collimation with short acquisitions we are able to obtain a true volume data set for evaluation of the venous system. Using the 3D display we can define vessel patency as well as determine early vessel encasement or invasion. The use of these display tools decreases the potential for false positive studies as well as indeterminate studies. Areas where the 3D display is especially helpful are at the confluence of the portal and superior mesenteric vein as well as the more distal portions of the portal vein. As surgeon become more aggressive in putting vascular grafts when only limited invasion is present the use of these 3D angiographic maps will become even more valuable.

Adenopathy

The presence of small nodes near the tumor bed or in the porta hepatis is not uncommon. Although CT is limited to detection of pathologic nodes by size criteria rather than by texture it is easier to detect small nodal masses with the higher image resolution provide by MDCT. Differentiation of small nodes from unopacified bowel loops is easier when good vascular opacification allows bowel enhancement thereby optimizing differentiation. Clusters of small nodes (1cm or less) that are infiltrated by tumor can often be suggested based on the CT appearance. However, unless specific size criteria or enhancement criteria are developed, even MDCT will have inherent limitations in the evaluation of nodal disease similar to Zeman et al. (28) who in using a TNM staging for pancreatic cancer found the nodal accuracy to be 58%. However, in the overall scheme the two observers used were able to correctly answer the question of resectable vs. nonresectable disease in 96% and 84% receptively.

Mesenteric and Omental Implants

Another cause of a false negative CT scan in the patient with pancreatic cancer is the ability to detect the presence of small omental and mesenteric implants. The use of MDCT with narrow collimation and close interscan spacing enhances the ability to detect smaller implants and tumor nodules than were previously possible. These tiny implants will often show some degree of enhancement, which improves our accuracy for detection. Implants can range in size but are often 5-10 mm in maximum cross sectional diameter. The review of CT as a volume should also impact on the ability to detect lesions that implant on bowel, the liver or mesentery.

Conclusion

Despite the extensive advancement s in our ability to image the pancreas as a result of innovations in scanner and workstation technology it is safe to assume that the progress will continue. During the next few years manufacturers will introduce scanners that acquire a minimum of 4 times as much information (up to 32 slices per second) with ever faster scan times. These scanners will be able to acquire data with .5 mm slice thickness routinely which will provide isotrophic data sets for CT angiography. Isotrophic data sets mean that the resolution is equal in any plane (x, y, or z axis) which will increase our ability to detect smaller lesions and more accurately stage the presence and extent of tumor. Workstations that are currently on proprietary hardware will move to more generic platforms with advanced graphic boards in place. These systems will have enhanced capabilities yet may cost 25% or less of what today’s systems typically cost. This will result in a wider diffusion of the technology beyond radiology into the operating room, surgical clinic and to the individual surgeon’s office and home. Although the future is hard to predict, it is likely that as network bandwidth increases and as hardware costs drop, many of today’s imaging paradigms may indeed continue to evolve, we are confident that these changes will continue to improve our diagnostic capabilities and therefore will improve patient care.

Table 1

Bibliography

1. Diehl SJ, Lehmann KJ, Sadick M, Lackmann R, Georgi M. Pancreatic Cancer: Value of Dual-Phase Helical CT in Assessing Resectability. Radiology 1998; 206:373-378.

2. Irie H, Honda H, Kaneko K, Kuroiwa T, Yoshimitsu K, Masuda K. Comparison of helical CT and MR imaging in detecting and staging small pancreatic adenocarcinoma. Abdom Imaging 1997; 22:429-433.

3. O’Malley ME, Boland GWL, Wood BJ, Fernandez-del-Castillo C, Warshaw AL, Mueller PR. Adenocarcinoma of the Head of the Pancreas: Determination of Surgical Unresectabiility with Thin-Section Pancreatic-Phase Helical CT. AJR 1999; 173:1513-1518.

4. Bonaldi VM, Bret PM, Atri M. Reinhold. A Comparison of Two Injection Protocols Using Helical and Dynamic Acquisitions in CT Examinations of the Pancreas. AJR 1996; 167:49-55.

5. Killius JS, Nelson RC. Logistic Advantages of Four Section Helical in the Abdomen and Pelvis. Abdom Imaging 2000; 25:643-650.

6. Hu H, He HD, Foley WD, Fox SH. Four Multidetector-Row Helical CT: Image Quality and Volume Coverage Speed. Radiology 2000; 215-55-62.

7. Chong M, Freeny PC, Schmiedl UP. Pancreatic Arterial Anatomy: depiction with Dual-Phase Helical CT. Radiology 1998; 208:537-542.

8. Foley WD, Mallisee TA, Hohenwalter MD, Wilson CR Quiroz FA, Taylor AJ. Multiphase Hepatic CT with a Multirow Detector CT Scanner. AJR 2000; 175:679-685.

9. Sim JS, Choi BI, Han JK, Chung MJ, Chung JW, Park JH, Han MC. Helical CT anatomy of pancreatic arteries. Abdom Imaging 1996; 21:517-521.

10. Novick SL, Fishman EK. Three-Dimensional CT Angiography of Pancreatic Carcinoma: Role in Staging Extent of Disease. AJR 1998; 170:139-143.

11. Johnson PT, Heath DG, Duckwall JR, Fishman EK. Enhanced Display of Vascular Anatomy with Stereoscopic Viewing. J Diagnostic Radiography and Imaging 1999; 2(1):25-28.

12. Calhoun PS, Kuszyk BS, Heath DG, Carley J, Fishman EK. Three Dimensional Volume Rendering of Spiral CT Data: Theory and Method. RadioGraphics 1999; 19:745-764.

13. Graf O, Boland GW, Warshaw AL, Fernandez-del-Castillo, Hahn PF, Mueller PR. Arterial Versus Portal Venous Helical CT for Revealing Pancreatic Adenocarcinoma: Conspicuity of Tumor and Critical Vascular Anatomy. AJR 1997; 169:119-123.

14. Choi BI, Chung MJ, Han JK, Han MC, Yoon YB. Detection of pancreatic adenocarcinoma: relative value of arterial and late phases of spiral CT. Abdom Imaging 1997; 22:199-203.

15. Bonaldi VM, Bret PM, Atri M, Reinhold C. Helical CT of the Pancreas: A Comparison of Cine Display and Film-Based Viewing. AJR 1998; 170:373-376.

16. Kuszyk BS, Bluemke DA, Urban BA. Portal-Phase Contrast Enhanced Helical CT for the Detection of Malignant Hepatic Tumors: Sensitivity Based on Comparison with Intraoperative and Pathologic Fndings. AJR 1996; 166(1):91-95.

17. Freeny PC. Radiologic Diagnosis and Staging of Pancreatic Ductal Adenocarcinoma. Radiol Clin North Am 1989; 27(1):121-128.

18. Freeny PC, Marks WM, Ryan JA, Traverso LW. Pancreatic Ductal Adenocarcinoma:Diagnosis and Staging with Dynamic CT. Radiology 1988; 166:125-133.

19. Bluemke DA, Cameron JL, Hruban RH, Pitt HA, Siegelman SS, Soyer P, Fishman EK. Potentially Resectable Pancreatic Adenocarcinoma: Spiral CT Assessment with Surgical and Pathologic Correlation. Radiology 1995; 197:381-385.

20. Raptopoulos V, Prassopoulos P, Chuttani R, McNicholas MMJ, McKee JD, Kressel HY. Multiplanar CT Pancreatography and Distal Cholangiography with Minimum Intensity Projections. Radiology 1998; 207:317-324.

21. Tabuchi T, Itoh K, Ohshio G et al. Tumor Staging of Pancreatic Adenocardinoma Using Earl and Late Phase Helical CT. AJR 1999; 173:375-380.

22. Lu DSK, Reber HA, Krasny RM, Kadell BM, Sayre J. Local Staging of Pancreatic Cancer: Criteria for Unresectability of Major Vessels as Revealed by Pancreatic-Phase, Thin-Section Helical CT. AJR 1997; 168:1439-1443.

23. Horton KM, Fishman EK. 3D CT Angiography of the Celiac and Superior Mesenteric Arteries with Multidetector CT Data Sets: Preliminary Observations. Abdom Imaging 2000; 25:523-525.

24. Fishman EK, Horton KM, Urban BA. Mulidetector CT Angiography in the Evaluation of Pancreatic Carcinoma: Preliminary Observations. J Comput Assist Tomogr 2000; 24(6):849-853.

25. Vedantham S, Lu DSK, Reber HA, Kadell B. Small Peripancreatic Veins: Improved Assessment in Pancreatic Cancer Patients Using Thin-Section Pancreatic Phase Helical CT. AJR 1998; 170:377-383.

26. Graf O, Boland GW, Kaufman JA, Warshaw AL, Fernandez-del-Castillo C, Mueller PR. Anatomic Variants of Mesenteric Veins: Depiction with Helical CT Venography. AJR 1997; 168:1209-1213.